General-purpose super-resolution models, particularly Vision Transformers, have achieved remarkable success but exhibit fundamental inefficiencies in common infrared imaging scenarios like surveillance and autonomous driving, which operate from fixed or nearly-static viewpoints. These models fail to exploit the strong, persistent spatial priors inherent in such scenes, leading to redundant learning and suboptimal performance. To address this, we propose the Regional Prior attention Transformer for infrared image Super-Resolution (RPT-SR), a novel architecture that explicitly encodes scene layout information into the attention mechanism. Our core contribution is a dual-token framework that fuses (1) learnable, regional prior tokens, which act as a persistent memory for the scene’s global structure, with (2) local tokens that capture the frame-specific content of the current input. By utilizing these tokens into an attention, our model allows the priors to dynamically modulate the local reconstruction process. Extensive experiments validate our approach. While most prior works focus on a single infrared band, we demonstrate the broad applicability and versatility of RPT-SR by establishing new state-of-the-art performance across diverse datasets covering both Long-Wave (LWIR) and Short-Wave (SWIR) spectra. Code is available at https://github.com/Yonsei-STL/RPT-SR.git.

In the Noisy Intermediate-Scale Quantum (NISQ) era, using variational quantum algorithms (VQAs) to solve optimization problems has become a key application. However, these algorithms face significant challenges, such as choosing an effective initial set of parameters and the limited quantum processing time that restricts the number of optimization iterations. In this study, we introduce a new framework for optimizing parameterized quantum circuits (PQCs) that employs a classical optimizer, inspired by Model-Agnostic Meta-Learning (MAML) technique. This approach aim to achieve better parameter initialization that ensures fast convergence. Our framework features a classical neural network, called Learner, which interacts with a PQC using the output of Learner as an initial parameter. During the pre-training phase, Learner is trained with the meta objective function. In the adaptation phase, the framework requires only a few PQC updates to converge to a more accurate value, while the learner remains unchanged. This method is highly adaptable and is effectively extended to various Hamiltonian optimization problems. We validate our approach through experiments, including distribution function mapping and optimization of the Heisenberg XYZ Hamiltonian. The result implies that the Learner successfully estimates initial parameters that generalize across the problem space, enabling fast adaptation.

Driven by the megatrend of aerial vehicles, there is an enormous increase in Unmanned Aerial Vehicles (UAVs) that opens up a new field of possibilities, applications, and services to the industry and public. The technological advancements along with the use of electronic control units in UAVs increase complexity. It requires high computing power, power management, and robust electrical and electronic architecture than they are currently used. There is no existing Electrical/Electronics (E/E) architecture standard for UAVs. In this paper, we design a novel E/E architecture for UAVs. The new E/E architecture for UAVs allows for the integration of more complicated and sophisticated functionalities inboard UAVs, including low battery consumption management, efficient navigation, high-performance, chassis, stability control, etc. We also presented future UAV E/E architecture based on zones for safe operations.

Learning adaptable policies in the absence of explicit reward signals is a challenging problem in reinforcement learning. We propose an algorithm that disentangles the morphology-aware variables from skill latent space for adapting quickly to unseen morphological changes. Through several task learning experiments using MuJoCo Ant environments, we demonstrate that the agent can perform zero-shot inference and adapt to mild modification in morphology within an expected performance range. Furthermore, in the case of severe unseen morphological damage to the agent’s body, with the help of add-on training steps, we can subjoin additional value incorporated into the disentangled latent space without catastrophically destroying the pre-trained network. We observe that the proposed separable-skill based method outperforms prior evolutionary meta-learning-based approaches, and the presented approach opens up research direction toward reinforcement learning for open-world novelty. Our source code is available at:https://github.com/boratw/sd4m.

Many significant research achievements in the past few decades have demonstrated that convolutional neural networks (CNNs) could be capable of steering wheel control which is the basic and essential maneuver of autonomous vehicles. We propose an end-to-end steering controller with CNN-based closed-loop feedback for autonomous vehicles that improves driving performance compared to traditional CNN-based approaches. This paper demonstrates that the proposed neural network, DAVE-2SKY, is able to learn to inference steering wheel angles for the lateral control of self-driving vehicles through initial supervised pre-training and subsequent reinforced closed-loop post-training with images from a camera mounted on the vehicle. We used the PreScan simulator and Caffe deep learning framework for training under diversified circumstances in a software in the loop (SIL) simulation environment. We used DRIVETMPX2 computer to implement a self-driving car for an experimental validation of the proposed end-to-end controller. The performance of the proposed system has been investigated under simulations and on-road tests as well. This work shows that the CNN-based end-to-end controller performs robust steering control even under partially observable road conditions, which indicates the possibility of fully self-driving vehicles controlled by CNN-based end-to-end steering controllers.

We propose a real-time near-optimal motion and trajectory planner using an interior-point method (IPM) based simultaneous dynamic optimization approach applicable to automatic parallel parking. Partitioning an automatic parking zone into two areas with and without inequality constraints drastically improves the convergence efficiency of the IPM-based simultaneous dynamic optimization of nonlinear programs while retaining the accuracy and stability of simulation for optimization. An optimal ready-to-reverse point (RRP), from where backward moving parking maneuver starts, is provided based on the cost function of the tracking time and additional delayed time of gear shifts caused by multiple maneuvers. The stability and accuracy of the proposed method was verified through simulation using AMPL with an IPOPT solver.

We have implemented a mixed reality telepresence platform providing a user experience (UX) of exchanging emotional expressions as well as information among a group of participants. The implemented system provides a platform to experience an immersive live scene through a Head-Mounted Display (HMD) and sensory information to a VR HMD user at a remote place. Moreover, the user at a remote place can share and exchange emotional expressions with other users at another remote location by using 360° cameras, environmental sensors compliant with MPEG-V, and a game cloud server combined with a technique of holographic display. We demonstrated that emotional expressions of an HMD worn participant were shared with a group of other participants in the remote place while watching a sports game on a big screen TV.

We proposed a hand-free Natural User Interface (NUI) for VR Head-Mounted-Display (HMD) with infra-red (IR) based sensor tracking facial gestures. We have realized NUI based on the real-time recognition of user intuitions for VR HMD without any additional control devices except for a built-in Gyroscope and IR couplers with readout circuitry integrated in the foam interface of an HMD. We implemented seven control commands affordable for 3D interactions with virtual objects. The experimental data show that the proposed system provides a convenient and efficient 2D/3D user interface for both manipulating objects and controlling commands while wearing a VR HMD.

A technique for matching both the eye convergence and accommodation in a fixed screen non-see through 3D VR HMD(3 Dimensional Virtual Reality Head mount Display) is presented. The user field test data show that significant improvement in dizziness problems as well as eye comfortability of users. The proposed method may provide a crew for resolution to solve simulator sickness caused by the binocular disparity of 3D VR HMD.

Immersive VR, a technology getting attention in recent years, is widely applied to the realm of serious games because it can provide users with both fun and intriguing experiences. This poster proposes a self-training VR serious game for fire evacuation drill with concurrent tele-collaboration among avatars controlled by and synchronized with multiple-users in remote places. We introduce a system architecture of both single user and its extension to multiple-user system. The single user system consists of wearable sensors and 3D VR HMD to synchronize a user's motions to one's own avatar in the virtual environment. We can easily extend the system to the multi-user mode through the Unity game cloud server. The multi-user mode enable players to experience a tele-existence so that they can collaborate in the virtual environment, and they can concurrently navigate while interacting with virtual objects as if they coexist in the same space.

We have proposed and implemented a hologram based tele-existence platform to provide user experience of emotional exchange among a group of participants in both real and virtual environments. The system provides not only immersive live scene to a user wearing VR HMD at a remote place, but also share and exchange emotional expressions with other users at other place by using a 360° camera and cloud server combined with a holographic technique. In the experiments, 6-different basic emotional expressions of user wearing VR HMD were transmitted, and replayed on the hologram display at the place where the group of participants were watching broadcasting of a sport through large screen TV.

We propose a steering control method for an autonomous vehicle by controlling the steering rate instead of the steering angle and describe a method for extracting the steering rate from the reference path using the relationship with curvature and slip angle. The proposed steering control method can be used not only for planning steering control input from the desired trajectories at the planning stage but also for the instantaneous planning of a driving path during vehicle motion. From the desired trajectory plotted in a two-dimensional XY coordinate system, the steering rate is extracted from the reference path, as a function of x. The extracted steering rate functions as the control input to the vehicle's controller for changing the front wheel's angle. MATLAB simulations of a double-lane-change maneuver were conducted for two scenarios, i.e., when the desired trajectory is given and when the driving path is planned instantaneously. For the double-lane-change maneuver, the curvature and the steering rate input to the controller are extracted from the virtually generated instantaneous reference path. The robustness of the proposed model and method are verified by re-constructing the trajectory traveled when the calculated steering rate is input into the controller.

An analog front-end circuit for 13.56MHz RFID interrogators compatible with ISO14443, ISO15693 and ISO18000-3 Mode 1 RFID interrogators was designed and fabricated by using 0.35μm double poly CMOS process. The fabricated chip was operated at 3.3 volt single supply. The results of this work can be provided as reusable IPs in a form of hard or firm IPs for designing single chip 13.56MHz RFID interrogators.